fromhttp://www.cisl.ucar.edu/info/FORMS/KNP1-6.html

Knowledge Networking (Abstract)

The Knowledge Networking (KN) initiative focuses on the integration of knowledge from different sources and domains across space and time. Modern computing and communications systems provide the infrastructure to send bits anywhere, anytime in mass quantities-radical connectivity. But connectivity alone cannot assure:

1. useful communication across disciplines, languages, cultures;
2. appropriate processing and integration of knowledge from different sources, domains, and non-text media;
3. efficacious activity and arrangements for teams, organizations, classrooms, or communities, working together over distance and time; or
4. deepening understanding of the ethical, legal, and social implications of new developments in connectivity.

Knowledge Networking

To "know" about something is a much stronger claim than to learn about it or to gather information on it. "Knowledge" implies consensual verification, as well as the ability to predict and shape outcomes. Advances in computing and communications now hold the promise of fundamentally accelerating the creation and distribution of information. However, the construction of knowledge requires more than collecting and transmitting large amounts of data. *Building knowledge requires the scientific community coming to grips with new forms of gathering data, new tools to manipulate and store information new ways of transforming that information, and new ways of working together over distance and time*. The challenge for NSF is to facilitate the evolution from today's emphasis on information and distributed data to emerging systems for knowledge and distributed intelligence. The payoff for the scientific community is that interdisciplinary communities that can be joined in sharing data, accumulating information and building knowledge together will treat complex problems, traditionally addressed within disciplinary boundaries. This shift from simple information access to knowledge networking holds great promise for to transforming society and science.

NSF represents large science, engineering, and education communities that understand and can contribute to building these knowledge networks. Technological advances, spurred by the NSF, now enable scientific practitioners, who may be widely dispersed, to become a science network, sharing and integrating data, analyzing information and synthesizing knowledge. NSF wants to expand and scale up these activities in the sciences and engineering, enabling society to apply similar strategies throughout its information infrastructures.

The NSF Knowledge Networking initiative creates a program of closely interconnected activities to facilitate advances enabled by simultaneous revolutions in technology, content, and epistemology. The intellectual insights and the new process of science enabled by knowledge networking are central to NSF's mission.

Challenges

One challenge is to support activities that will create new ways of collecting, transforming, representing, sharing, and using information. The support must be applied effectively across a wide range of activities to enable the solution of the Knowledge Networking challenge and to provide to scientists, engineers, and society useful and easily implemented solutions to complex problems.

A complementary challenge is to comprehend the human dimensions associated with knowledge networking communities. Multidisciplinary knowledge networking efforts will fail unless we understand and provide for the learning environments that enable skill sets, conceptual models, and values to be rapidly shared across disparate fields.

Goals

Designing tools for gathering and analyzing data. New types of tools are required to collect, share, and manipulate increasingly complex data sets and structures. Utilizing these tools involves innovations in computing, advances in telecommunications, and the development of more sophisticated algorithms and hardware/software systems.

Building the next generation of representations. Data, information and knowledge require increasingly complex representations. New kinds of media are required to enable the communications of new types of messages and meanings. For example, transforming symbolic information into sensory form (e.g., visualization) necessitates translating scientific and mathematical notations into tangible modalities.

Extending the human infrastructure that underlies knowledge networking. Generating new ideas increasingly involves participants in knowledge networks communicating with one another in real time and obtaining data from disaggregated sources. Expanding the knowledge networking community to new participants requires:

Mastering a common language and a generally accepted set of theories and conceptual models (to provide a framework for communication) Inculcating communally defined processes of collecting and analyzing data (to enable sharing and validating information) Developing proficiency in design, reasoning, and argumentation (to facilitate the evolution of ideas) Accepting a common set of values that include respect for others' perspectives and for intellectual property (to encourage wide participation) Strategies

In keeping with NSF's strategic plan, the knowledge networking initiative proposes three research strategies.

Maximizing cross-disciplinary research to make use of different needs and demands of distinct research communities; Addressing the common problems shared by different research communities; Leveraging existing research activities and building upon them, rather than starting from scratch. NSF also envisions several strategies to launch and manage knowledge networking: Creating ongoing working groups to integrate disciplinary issues; Integrate community input through workshops; Promote interdisciplinary research; Involve all of the Directorates at NSF. NSF as a Catalyst

NSF can serve as a catalyst for creating knowledge networks. Part of this role involves supporting the development of enabling technologies (infrastructure), such as new algorithms and software systems; data structures; metadata; standards for interoperability, communications links, and computational platforms. These enhancements extend from innovative processes that bring researchers together in distributed collaboratories, to tools for analyzing and interpreting data in new ways, to sophisticated learning environments that help participants discover and integrate new knowledge.

The other portion of this catalytic role involves conceptualizing knowledge networking as collective action among scientific communities ranging across many fields and disciplines. By sharing disparate data and diverse perspectives, a community develops a common, evolving understanding of a complex topic. As the community's conception of the issues expands and deepens, its membership grows to include participants with new perspectives and backgrounds. Given its long experience with how this process of acculturation and distributed intelligence occurs in the scientific enterprise, NSF is positioned to aid in the development of the technological infrastructures, collaborative activities, and human communities needed for knowledge networking across in society as a whole.

The Knowledge Networking Initiative Integrates Layers of Achievement

The Knowledge Networking Initiative aims to create the underlying science and the tools, infrastructure, and distributed intellectual processes to achieve the layered aims shown in Figure 1.

The overarching goal is improving our understanding of and ability to manage larger and more complex natural, social, and material phenomena. Knowledge networking can enhance the operations of many human enterprises, with science and education the most obviously relevant to NSF's mission. The crucial added benefits that knowledge networking brings to the scientific enterprise are the abilities to:

Couple models, knowledge, data, instruments, and intellectual activity across space, time, and disciplinary boundaries, Work with new types of content and knowledge bases of radically increased scope and scale, Enhance the overall cognitive ecology of science and engineering. Achieving these aims of coupling, scope, and intellectual community depends critically upon new levels of functionality in information infrastructures. We need a better understanding of how to push or pull relevant information wherever, whenever and to whomever it is useful; how to create true semantic interoperability in heterogeneous knowledge environments; and how to make knowledge maximally accessible with new modalities of interaction such as real-time multimedia, visualization, and simulation.

Achieving such new functionality and making them widespread and universally accessible also requires re-conceptualizing the human processes involved in creating and disseminating knowledge. The groups involved include data gathering enterprises such as field research teams, observatories, and cyclotron facilities; information transmission functions such as messaging, publishing, and library systems; and integrating/stabilizing infrastructures such as standards and user groups. Each type of human interaction in the overall scientific process must alter if knowledge networking is to reach its full potential. Figure 2 and Figure 3 capture more of the dimensions of Knowledge Networks, and communicate their dynamic nature. Examples of Social Objectives and Relevance Outcomes that could be enabled by the Knowledge Nets Initiative.

The following examples are potential outcomes of Knowledge Networks. They illustrate the use of science and technologies to meet larger societal goals. These examples involve science and the use of scientific information that are possible only with the use of Knowledge Networks. In addition the examples require advancements in one or more of the subsystems (such as social use of the new knowledge, science modeling, datamelding, and real-time networks) from each of the top three layers of the Framework discussed above.

Coping with Natural Disasters

In 1995, twelve forest fire fighters died tragically when they were trapped on the side of a mountain in Colorado, unaware that sudden changes in meteorological conditions had caused a change in the path of the fire. Although some data were available indicating a shift in the fire, this information could not be delivered to the scene of the fire in a timely and clearly understood manner. The enterprise, infrastructure and tools which constitute the framework of the Knowledge Nets (KN) initiative will enable an integrated framework which does not exist today for dealing with natural disasters, ultimately leading to minimizing loss. Specifically, the KN initiative could support the development of coupled fire and atmospheric models. These models require as input detailed knowledge of topography, ground cover and synoptic weather conditions. These data exist in various data bases spread over the country and are expressed in different formats. The result from a simulation must be overlaid with the detailed knowledge of the location of human and physical resources. In cases where fire is near more populated areas, as in the Oakland, California fires, additional information about the demographics and civil infrastructure must be incorporated. Even if this synthesis of rapidly changing information could be assembled today, delivery to strategic locations in an understandable form would still be necessary to ensure benefit. The infrastructure and tools components on the KN initiative are "glue" that will enable the effective management of natural disaster situations.

Aviation Safety:

Delivery of current information to the cockpit and proper pilot training are essential elements in improving air safety and reducing operating costs. Significant progress has been made in pilot training and alerting pilots to potential life threatening situations. Examples of improved safety and reduced operating costs that could result from the research supported by KN are: 1) At many airports in the US, information on low-level wind shear coupled with Doppler radar allows air traffic controllers to alert pilots to unusual meteorological conditions. Improvements to the current capabilities could save additional lives and money for the airline industry. The current information that is assembled by air traffic controllers is of limited predictive value and must be reduced to a few numbers to allow the pilot to comprehend the information in the cockpit during takeoffs and landings. Synthesizing the results of models of the atmosphere and air traffic into the cockpit and control towers would allow the pilot and controller to better prepare for approaches or takeoffs through in-flight simulation of conditions. In addition to offering improved safety, this information will save significant fuel costs because planes would not have to be routed to different approaches at the last minute due to changed conditions on the ground. 2) The FAA is considering the feasibility of free flight by commercial airlines. This concept would allow aircraft to take the most direct route between cities rather than following established routes that pass predetermined checkpoints. Essential for free-flight are current information on weather conditions, location of other aircraft and conditions at airports along the route. Gathering this information and synthesizing and delivering it in a useful form is beyond our current capability. The airline industry estimates the annual savings, which may be recognized by implementing free flight, is tens of millions of dollars. 3) In-flight icing conditions are difficult to detect and even more difficult to predict. Several recent airline disasters have been attributed to icing. Improvements in the detection, prediction and delivery systems available to the airline industry are necessary to overcome this silent threat. The enterprise, infrastructure and tools that will be developed as part of the KN initiative will accelerate the ongoing research into in-flight icing.

Monitoring and Restoring Landscape Change: the Florida Everglades

Large scale human impacts on landscapes have complex biological, social and economic consequences. The dramatic impact human activities have had on ecosystem function in the Everglades has elicited an enormous amount of research, land and water management and conservation activities. The health and recovery of the Everglades and adjoining areas is now being considered by stakeholders in many sectors: several federal agencies (DOI is dispersing $200 million for restoration), scores of state agencies and local jurisdictions, hundreds of research activities, academic centers from public and private universities, the sugar industry, two tribal nations, along with conservation and public grass-roots organizations. The dynamics of the interactions between all of these parties creates knowledge chaos conditions which leads to duplication of effort, gridlock, turf conflicts, organizational and political uncertainty, needless competition and distrust between stakeholders.

Information inputs for rational Everglades planning and recovery come from highly-distributed and diverse sources such as from long-term biological surveys, hydrochemical monitoring, watershed flow models, land use change analyses, remote sensing, economic analyses and human demographic studies, among inputs from other research, sociological and economic areas.

The accumulation, representation and communication of such rich and heterogeneous of data sets that span: decades of time, numerous research disciplines and diverse stakeholders in multiple sectors of the economy, creates enormous challenges and equally enormous potential payoffs for effective knowledge networking. Research and infrastructure development on data integration, data mining, geo-spatial visualization, human interactions, network communications, data sharing, the coordination of long term monitoring, all within the context of a well-defined, nationally important, environmental effort would have immediate value to society and represents an immediate payoff test-bed for new knowledge networking approaches. Examples of How Knowledge Networking can Facilitate Research and Education

NetCDF: A Tool That Facilities Collaboration

Though a typewritten table of numbers once sufficed to characterize most quantitative studies, scientific efforts now often yield quantities of data that can be structured, interpreted, and utilized for further study only by computer. Thus, methods for inter-computer data exchange represent critical infrastructure for scientific collaboration. Though there are common means for transferring human-readable material and for selecting and retrieving information from data-base management systems, there is little agreement on methods to convey some of the most common data structures, such as vectors and multidimensional arrays, used in certain disciplines.

To help address this problem, Unidata developed the Network Common Data Form (NetCDF). The method is not for end-user; rather, it is a programmer's tool kit for storing and retrieving data in files that are portable (i.e. transferable between dissimilar computers) and self describing (i.e. that contain enough information to obviate the need for ancillary documents on dimensionally, variable names, units of measure, etc.). The NetCDF development represents, on a very limited scale, a harbinger of impact KN will have on science, viz. the creation of enabling technologies that will result in the generation of fundamental new knowledge.

The NetCDF's existence and free availability appears to have had a positive effect on collaboration as well as on the development of scientific software. Hundreds of commercial and non-commercial organizations all around the world and representing a wide variety of disciplines have adopted NetCDF for scientific analysis or visualization.

Tools for Capturing, Validating, and Sharing Mathematical Knowledge

Mathematicians are attempting to build computer systems that effectively represent mathematical knowledge, and that enable the construction of databases of mathematical results and mechanically checked proofs in forms that are readable and usable by people. Users of such systems could shift among different but clear and unambiguous representational syntaxes that capture the same underlying mathematical knowledge in forms that are tailored for use in different contexts. Users of such systems could know that every mathematical result represented has a proof that has been checked by computer and is available for inspection. Mathematicians who have discovered new results may wish to add them to the collection, helping to publicize these results, certify their validity, and increase the overall capability. Such tools may also help to systematize mathematical knowledge and to integrate theoretical knowledge with computation. Such systems could be a clear example of the power of Knowledge Networking for dealing with organized knowledge rather than isolated facts, and the broad utility of mathematical knowledge makes it a compelling candidate. Wide availability of such capability and knowledge through integration with the Internet could lead to new generations of researchers, teachers, and students using such tools routinely in any context where mathematics is used. Research Opportunities in Knowledge Networking:

Knowledge Networking presents a number of research challenges and opportunities. These can be organized under a set of topics or threads, which we have termed *Interactivity, Representation, Cognition, Agents, Corpora*.

Interactivity

Interactivity research studies the creation and maintenance of dynamic, content-rich relationships among people, instruments & tools, data, and artificial agents, using multiple modalities. Technologies that enable such interactivity encompass input/output devices, communication networks, and their interface characteristics, adapted with the aim of making the best match to what is known about the needs and requirements of individual people, groups, teams, and organizations for effective interaction. The critical multidisciplinary aspects of Interactivity research result from the need to uncover common foundations for understanding widely differing types of participants (e.g. people or agents with particular skills; specialized instruments) coupled through unique domain-specific activities (e.g. doing geoscience or doing disaster relief) integrating problem- and domain-specific information (e.g., specialized datasets or knowledge bases), via a variety of media and channels (text, video, etc.), under a range of specific constraints (e.g. quality-of-service; sensory limitation such as no vision or hearing, etc.). Another multidisciplinary driver is the need to understand how to apply the fruits of Interactivity research effectively in many different domains. New interdisciplinary Knowledge Networking research under the Interactivity thread includes:

• Access and "universal access"
• Body language and facial expression capture/generation
• Dialogue and discourse structures and constraints
• Ecological & virtual-environmental interactions
• Image processing and gesture recognition
• Intermodal mappings for people with disabilities
• Haptics, smell, taste and balance technologies
• Multimodal interactivity
• Network and communications infrastructures for interactivity
• Remote access/control and teleoperation
• Signal processing and understanding
• Speech recognition and natural language understanding
• Real-time and time-modified interactivity
• Valuation and incentives for interactivity

Representation

Research on representation studies the processes through which participants (people, groups, agents, etc.) model and encode knowledge about entities, processes, or phenomena in particular representational media, and, conversely, reconstruct meanings and semantics for representations in their contexts of use.

The critical multidisciplinary aspects of Representation research result from the need to uncover common foundations for understanding how widely differing types of participants (e.g. people or agents with particular domain- or culture-specific viewpoints; specialized data-gathering instruments), represent problem- and domain-specific entities or processes (e.g., protein molecules; organizational workflows), of differing representational level (e.g., sensory; cognitive), scale and complexity, for use in unique domain-specific activities (e.g. doing bioscience or doing collaborative design) via a variety of representational media and modalities (text, software, graphical data, simulations, in visual, audio, haptic modalities, etc.), under a range of specific constraints (e.g. size limitations, specificity constraints). Another multidisciplinary driver is the need to understand how to apply the fruits of Representation research effectively in many different domains. New interdisciplinary Knowledge Networking research under the Representation thread includes: Representation of new entities or attributes, such as:

• Complex data types
• Complex systems and their structure
• Domain- and discipline-specific objects, actions, and processes
• Gestures and facial expressions
• Human sensory information: touch, smell, taste, and balance
• Illumination and rendering
• Know-how and commonsense knowledge
• Large-scale systems and phenomena
• Mathematics and logic
• Ontologies
• Open physical, computational and biological systems
• Organizational processes and workflows
• Real world objects, processes and environments
• Scientific principles, methods, and theories
• Uncertainty
• Complex operations on representations, such as:
• Automatic generation of representations
• Domain-independent abstraction (e.g., of texts and images)
• Compression
• Integration, fusion and interoperability;
• Interpretation of representations in context
• Multimedia indexing, abstraction
• Translation of representations
• New representational techniques and media such as:
• Distributed representations
• Distributed interpretations
• Non-digital representational media such as culture, the physical world, molecules and Biological objects (e.g., DNA)
• Multiperspective representations
• Simulations and computer models
• Tools for joint or collaborative knowledge construction and representation
• New uses for representations
• Cognition
• Cognition research investigates interlinked processes of perception, reasoning, memory, learning, and action by participants in physical and socio-cultural situations.
• The critical multidisciplinary aspects of Cognition research result from the need to uncover common foundations for integrated understanding of all phases of cognition, as carried out by a wide variety of cognitive entities (e.g., people, artificial agents, groups, organizations), cognizing (perceiving, reasoning/learning about, acting with) domain-specific phenomena of differing character, scale and complexity (e.g. perception of surface textures; organizational memory), in a wide variety of physical and social contexts (e.g. a laboratory, a crisis management scenario) under a range of specific constraints (e.g. complexity, realizability, or real-time constraints).
• Another multidisciplinary driver is the need to understand how to apply the fruits of Cognition research effectively in many different domains.
• New interdisciplinary Knowledge Networking research under the Representation thread includes:
• New cognizing entities:
• Cognition by individuals, groups, teams and organizations
• Empirical studies of Knowledge networks as arenas for scientific experimentation, data gathering, analysis
• Human comprehension in networked environments
• Organizational and community memory systems
• New objects of cognition such as:
• Behavior and event processing of non-rigid objects
• Complex, distributed, and open systems

Groups, teams, organizations, institutions Specific domain entities Tasks of high complexity New cognitive issues and methods: Distributed cognition Dynamic adaptation Error processing and propagation; Exploiting parallel architectures for computation; Focus of attention High-level reasoning; Knowledge-based information processing; Learning effects of human exposure to virtual and real environments. Non-conceptual cognition; Perceptual, motor, and sensory-motor models; Perception-based problem solving Realizability of cognitive models Situated cognition Symbolic and geometric processing; Transfer of learning; skill training and acquisition Cognitive aspects of trust and believability

Agents

Agents research studies the active and sometimes physically embodied algorithms, software, communications, and tools that can assist people in Knowledge Networking activities. Examples of agents include knowledge agents that seek and manipulate specific data or information collections ("Knowbots") from interconnected commuter networks, and cooperative physical agents such as robots, intelligent devices, special instruments, and other non-human natural agents or environments. The critical multidisciplinary aspects of Agents research result from the need to uncover common foundations for understanding how to support and augment a variety of people, teams, groups, and organizations, each with particular domain- or culture-specific needs, in performing unique domain-specific activities (e.g. doing bioscience, doing collaborative design, or emergency management), using a varied array of resources (scientific data sets, distributed simulations, specialized instruments), under a range of specific constraints (e.g. time, methodological, or performance quality constraints). Another multidisciplinary driver is the need to understand how to apply the fruits of Agents research effectively in many different domains. New interdisciplinary Knowledge Networking research under the Agents thread includes: Autonomy Building (and dismantling) information-rich virtual communities and organizations Domain-specific contextual knowledge and commonsense knowledge for agents Coordination of activities and knowledge in heterogeneous systems and environments Degrees and types of augmentation and support for participants such as people, teams, or organizations. Designs and criteria for sensory-motor systems Distributed control Domain-specific contextual knowledge and commonsense knowledge for agents Dynamic adaptation and evolution of agents Engineering methodologies Interoperability of agents Incentive structures Knowledge at new scales: large collections of tiny, heterogeneous distributed agents; distributed knowledge networks for control of MEMS Load and complexity management Mathematical algorithms, machine architectures, and networking technologies for knowledge agents in information spaces Multi-agent systems Modularity, parallelism and complexity Pathologies and immune systems in large-scale human-computing aggregates, e.g. malicious agents, viruses, junk, "knowledge storms" Possible or optimal domain, range and scope of the agents' functionality Principles of decomposition and organization of tasks and resources (division of labor) Robustness, fault-tolerance, and reliability Specific domain-dependent agents for assisting in information analysis, decision making, and remote control of instruments and access to information resources Trust, confidence, and believability User, team, and organizational requirements and their evolution.

Corpora

Investigations of corpora (plural of "corpus") research the entire lifecycle (creation, structuring, storage, maintenance, use and disposal) of general and community-specific collections of data, information, and knowledge, ranging across ad hoc data collections, complex scientific databases, large and distributed digital libraries and even such unconventional entities as digital forms of artifacts in museums. Research in Corpora is a critical enabler of Knowledge Networking: people's ability to access, retrieve and comprehend information from complex databases and sources depends on how that information is created, structured, stored, presented, and managed. New interdisciplinary Knowledge Networking research under the Corpora thread has two objectives: To accelerate cross-disciplinary database research, and to develop new kinds of cross-disciplinary data-sharing mechanisms, infrastructures, and relationships that can facilitate new interdisciplinary experimental research. Relevant research topics include:

• Active and real-time databases and data sources
• Classification and taxonomizing processes
• Collection and indexing of retrospective and real-time data sources
• Community, organizational, and social filtering of information and knowledge
• Data and knowledge mining
• Dealing with evolution of structure, function, content, and user requirements
• Digital libraries and repositories across disciplines and application domains
• Disciplinary databases; multi-lingual, cross-cultural, societal collections
• Dissemination and distribution
• Domain-specific taxonomies and ontologies
• Dynamic synthesis of new structure, views, and metadata
• Economics and valuation of information and content
• Evolution of structure, function and content
• Experimental evaluation
• High-confidence and reliability
• Intellectual property rights and ownership
• Intelligent transaction processing
• Meta-data research
• Multi-modal information management
• Object and multimedia environments
• Organizations as active knowledgebases
• Public access policies
• Reliability, security, quality of data and data services
• Searching, filtering, fusion, indexing and retrieval tools
• Security and authorization
• Structure, functionality and organization of corpora
• Technologies of intellectual property, e.g., "terms and conditions"
• Vertically and horizontally linked data sources